Journal paper on acoustics of breath noises online

I am very happy to announce that the journal paper I first-authored called Acoustics of Breath Noises in Human Speech: Descriptive and Three-Dimensional Modeling Approaches is now online and open access via ASHA and PubMed. My co-authors for this project were Susanne Fuchs (ZAS Berlin), Jürgen Trouvain, Bernd Möbius (both UdS), Steffen Kürbis, and Peter Birkholz (both TU Dresden).

The paper gives the first description of the spectral characteristics of speech breath noises produced by a large number of (German) speakers. In addition, we modeled in- and exhalations using 3D-printed vocal tract models, that were produced using MRI. The main findings are:

1. Breath noises have several weak peaks that align with resonances found in a very controlled setting where participants inhaled with the VT configuration of a central vowel in Hanna et al. 2018.
2. Comparing in- and exhalation spectra in the 3D-printed VT models, airflow direction changes the spectral properties of /s ʃ ç i:/, but not of the other sounds we investigated.
3. We tried to compare real inhalations and model inhalations but there is a myriad of mechanisms that are either hard to model or still un(der)-researched for speech breathing, so there's many interesting things left to do in that field.

The findings may help with the automatic detection of pathologies such as COVID-19, pathological cries and coughs in infancy, or vocal fold paralysis, or the emotional or cognitive state of the speaker. I hope that this paves the way for further work using the acoustics of breath noises in general, but also in pathological and synthesized speech.